*For correspondence: rosbash@

brandeis.edu

Competing interests: The

authors declare that no

competing interests exist.

Funding: See page 18

Received: 24 July 2016

Accepted: 01 December 2016

Published: 09 December 2016

Reviewing editor: Hugo J

Bellen, Baylor College of

Medicine, United States

Copyright Chen et al. This

article is distributed under the

terms of the Creative Commons

Attribution License, which

permits unrestricted use and

redistribution provided that the

original author and source are

credited.

Genome-wide identification of neuronalactivity-regulated genes in DrosophilaXiao Chen1,2, Reazur Rahman1,2, Fang Guo1,2, Michael Rosbash1,2*

1Howard Hughes Medical Institute, Brandeis University, Waltham, United States;2National Center for Behavioral Genomics, Department of Biology, BrandeisUniversity, Waltham, United States

Abstract Activity-regulated genes (ARGs) are important for neuronal functions like long-term

memory and are well-characterized in mammals but poorly studied in other model organisms like

Drosophila. Here we stimulated fly neurons with different paradigms and identified ARGs using

high-throughput sequencing from brains as well as from sorted neurons: they included a narrow set

of circadian neurons as well as dopaminergic neurons. Surprisingly, many ARGs are specific to the

stimulation paradigm and very specific to neuron type. In addition and unlike mammalian

immediate early genes (IEGs), fly ARGs do not have short gene lengths and are less enriched for

transcription factor function. Chromatin assays using ATAC-sequencing show that the transcription

start sites (TSS) of ARGs do not change with neural firing but are already accessible prior to

stimulation. Lastly based on binding site enrichment in ARGs, we identified transcription factor

mediators of firing and created neuronal activity reporters.

DOI: 10.7554/eLife.19942.001

IntroductionUpon stimulation, immediate-early genes (IEGs) are induced rapidly and transiently without de novo

protein synthesis in mammalian neurons (Fowler et al., 2011). This activity-dependent gene regula-

tion is important as it affects the ability of an animal to convert transient stimuli into long-term

changes (Huh et al., 2000; Kandel, 2001; Nestler, 2001; Spitzer et al., 2000). Hundreds of IEGs,

including those specific to tissues and stimulation paradigms, have been identified in mammals since

the discovery of the first neuronal IEG, c-fos, by Greenberg and his colleagues in 1986

(Greenberg et al., 1986; Herschman, 1991; Spiegel et al., 2014). Expression of these IEGs are

induced within an hour, usually with big amplitudes (over 10 fold), and many IEGs are shared

between different types of neurons (Spiegel et al., 2014). Functionally, these IEGs are highly

enriched for transcription factors, which subsequently trigger a secondary transcriptional response

(Spiegel et al., 2014; West and Greenberg, 2011). The secondary response genes (SRGs) in con-

trast take longer to induce (a typical assay is 6 hr post-stimulation), are involved in many different

processes, are more neuron-specific and function at least in part to promote neuron survival, den-

dritic morphogenesis and regulate synapse formation (Bloodgood et al., 2013; Hong et al., 2008;

Lin et al., 2008).

Not all IEGs respond to all types of stimulations, even the most robust ones like c-fos, Egr1 and

Arc (Bepari et al., 2012; Fields et al., 1997). For example, different stimulation paradigm-depen-

dent Ca2+ entry routes initiate different downstream pathways and lead to induction of distinct IEGs

(West and Greenberg, 2011). Stimulations other than neural firing like growth factors also induce

IEG expression, many of which are the same as those induced with neural firing (Jones et al., 1988;

Tullai et al., 2007). Depending on the induction dynamics, Tullai et al. divided the platelet-derived

growth factor (PDGF)-induced IEGs in human T98G glioblastoma cells into primary response genes

(PRGs) and delayed response genes (DRGs). Dramatic differences were shown between the two

Chen et al. eLife 2016;5:e19942. DOI: 10.7554/eLife.19942 1 of 21

RESEARCH ARTICLE

categories in terms of their functions, gene lengths, chromatin accessibility and RNA polymerase II

occupancy at promoter regions (Fowler et al., 2011; Kim et al., 2010; Tullai et al., 2007). PRGs are

usually optimized for rapid induction (such as shorter gene length and more permissive chromatin at

promoters), whereas DRGs are not different from other genes in the genome. Notably, PRGs and

DRGs are induced independent of protein-synthesis (‘cycloheximide-insensitive’), whereas SRGs are

protein-synthesis dependent (‘cycloheximide-sensitive’).

IEGs, or activity-regulated genes (ARGs; they are defined here as induced rapidly with neuronal

activity, i.e., mostly within an hour, but without regard to de novo protein synthesis) are poorly

defined in organisms other than mammals. Only three genes to date have been identified as

responding to increased neural activity in insects: kakusei (in honey bee), stripe (abbreviated as sr, in

honey bee) and hr38 (in silkmoth and the fruit fly Drosophila melanogaster) (Fujita et al., 2013;

Kiya et al., 2007; Lutz and Robinson, 2013). Moreover, only one genome-wide study using micro-

arrays had been performed more than a decade ago to identify seizure-induced ARGs from fly heads

(Guan et al., 2005). More high-throughput studies on Drosophila ARGs are therefore needed, not

only for identification and to provide mechanistic insight but also to design new tools to serve as

indicators of neuronal activity.

Here, we identified ARGs in a genome-wide manner, in fly brains as well as in sorted neurons;

they included dopaminergic neurons (DA) and a subset of circadian-related neurons (PDF+ neurons).

Fly ARGs vary with the individual stimulation paradigm and are surprisingly cell type-specific. Fly

ARGs are also more functionally diverse and have longer gene lengths compared to mammalian

ARGs. Chromatin at transcription start sites (TSS) of fly ARGs is more accessible at baseline than

other expressed genes but does not change with stimulation. Lastly, we used bioinformatics to iden-

tify key transcription factors that mediate ARG activation. Based on these factors, we generated

novel luciferase reporters for in vivo monitoring of neuronal firing.

Results

Genome-wide identification of firing-induced ARGs in fly brainsTo identify ARGs in fly neurons in a genome-wide manner, the pan-neuronal driver Elav-GAL4 was

used to drive expression of UAS-ChR2-XXL, to artificially fire neurons by illuminating flies with a blue

LED (Dawydow et al., 2014). A 30 s 10 Hz LED exposure was sufficient to induce a uniform seizure

within seconds, and all flies were able to recover within 15 min. Fly brains were dissected either

before (0 min), or 15, 30, 60 min after stimulation (Figure 1A). Flies expressing only Elav-GAL4 were

illuminated in parallel and used as a control strain. RNA was then extracted from these samples and

made into mRNA libraries for deep sequencing.

96 genes show significant increases in 60 min with p-value < 0.01 of both exact test and general

linear model (GLM) in UAS-ChR2-XXL flies (N = 3 biological replicas); very few genes show significant

decreases (Figure 1B). Ranking #1 and #2 are the previously identified insect IEGs hr38 and sr, with

a 90- and a 10- fold increase respectively, suggesting the stimulation paradigm is sufficient to fire

neurons in the brain (Figure 1C, Figure 1—source data 1). The 30 s LED exposure affected many

fewer genes in control flies: four were up-regulated, and three down-regulated with p-values<0.01

by both statistical tests (Figure 1—figure supplement 1A, N = 3). 3 of the four up-regulated genes,

CG13055, hsp23 and hsromega, had induction amplitudes statistically indistinguishable from those

of the UAS-ChR2-XXL flies and are therefore labeled as light-induced genes in Figure 1—source

data 1 (Figure 1—figure supplement 1B). However, they were not removed from the ARG list

because light can induce neuronal firing and some of these genes are also induced by other stimula-

tion paradigms (see below).

Gene Ontology (GO) analysis using GOrilla (http://cbl-gorilla.cs.technion.ac.il/) on the 96 ARG

genes shows that polII transcription factor activity is the most enriched function with a P-value of

1.29E-4 computed according to the mHG or HG model, followed by steroid hormone receptor activ-

ity and MAP kinase activity (Figure 1D). Several of these induced transcription factors encode pro-

teins with high sequence identity to well-known mammalian IEGs: HR38 shows 67% identity to

nuclear receptor subfamily four group A member 2 (NR4A2); SR has 82% identity to early growth

response protein 3 (EGR3); CBT is 61% identical to Krueppel-like factor 11 (KLF11). Taken together,

Chen et al. eLife 2016;5:e19942. DOI: 10.7554/eLife.19942 2 of 21

Research article Neuroscience

firing neurons with ChR2-XXL identified known as well as novel ARGs in flies, and the enrichment of

transcription factors in ARGs suggests some mechanistic similarity between flies and mammals.

Gene sizes of ARGs correlate with induction ratesTo assess the kinetics of ARG induction and to provide a global view, a heat map across time points

was generated (Figure 2A). The levels of most ARGs continue to increase until the last time point

assayed (60 min); only a few transcripts like cbt (2.6 kb) reach maximum expression levels at 30 min

(Figure 2B). Since shorter gene length enables the rapid induction of IEGs in mammals (Tullai et al.,

2007), we examined the gene lengths of fly ARGs. Surprisingly, the average ARG gene is longer

than that of all annotated genes as well as of all genes expressed in neurons (Figure 2C, gene anno-

tation obtained from FlyBase (ftp://ftp.flybase.net/genomes/dmel/dmel_r5.2_FB2007_01/fasta)). This

is true even for some of the most robust ARGs; hr38 is ~31 kb long, sr spans ~ 11 kb, CG8910 is ~21

Figure 1. High-throughput sequencing and optogenetics reveal ARGs in Drosophila. (A) Stimulation paradigm for Elav-GAL4;UAS-ChR2-XXL flies. (B)

Volcano plots of individual genes (FPKM > 0) with log2 fold changes (x-axis) against p-value (y-axis, exact test). Genes significantly induced (p-

value < 0.01 for both exact test and GLM) are in red. The smaller area with grey background is the zoom-in view with larger scales. N = 3 biological

replicates. (C) Gene expression of hr38 and sr in Elav-GAL4;UAS-ChR2-XXL and Elav-GAL4/+ (control) flies 0 or 60 min after LED stimulation. N = 3

biological replicates, error bars represent ±SEM, n.s. represents non-significant. *p<0.05, ****p<0.0001, exact test. (D) Gene ontology analysis on ChR2-

XXL-induced ARGs. P-value is the enrichment p-value, not corrected for multiple testing. FDR q-value is the corrected p-value for multiple testing using

the Benjamini and Hochberg (1995) method (Eden et al., 2009; Eden et al., 2007). Number of genes indicates the number of ARGs with the specified

functions.

DOI: 10.7554/eLife.19942.002

The following source data and figure supplement are available for figure 1:

Source data 1. ChR2-XXL-induced ARGs in fly brains.

DOI: 10.7554/eLife.19942.003

Figure supplement 1. Light-induced gene expression.

DOI: 10.7554/eLife.19942.004

Chen et al. eLife 2016;5:e19942. DOI: 10.7554/eLife.19942 3 of 21

Research article Neuroscience

Figure 2. Gene sizes of ARGs correlate with induction rates. (A) Heat map of ChR2-XXL-induced ARG expression from 0, 15, 30 to 60 min (normalized

to 0 min expression) post-LED stimultaion. Each row represents one gene. Data are averages of 2 biological replicates. (B) Gene induction profiles of

cbt (gene length = 2.6 kb, red) and sr (gene length = 11 kb, blue) with LED stimulation in Elav-GAL4;UAS-ChR2-XXL flies (N = 2 biological replicates,

abbreviated as rep1/2). (C) Gene length distribution of all annotated genes in the genome (blue), genes expressed in brains (gray, FPKM > 0 at

Figure 2 continued on next page

Chen et al. eLife 2016;5:e19942. DOI: 10.7554/eLife.19942 4 of 21

Research article Neuroscience

kb and CG11221 is ~13 kb (Figure 2—figure supplement 1). In contrast, the canonical mammalian

IEGs c-fos and c-jun are only ~ 3.4 kb and ~3.2 kb respectively. This is despite the fact that the aver-

age gene size of flies (~9.3 kb) is considerably smaller than that of mammals (~47 kb for mouse).

To test if gene length affects ARG induction rates, we compared the fold changes of gene

expression between 0 and 15 min (FC15/0) relative to gene lengths. There is a weak anti-correlation,

suggesting that longer genes are indeed more slowly induced on average (Figure 2D). This is not

because overall induction amplitudes are smaller for longer genes, since the correlation disappears if

fold changes between 0 and 60 min (FC60/0) are compared to gene lengths (Figure 2E).

To gain more insight into these dynamics, nascent RNA and mRNA libraries were constructed in

parallel from fly heads using the same stimulation paradigm. Heads rather than brains were used

because of technical considerations: the much larger amounts of starting material required for

nascent RNA preparation are difficult to obtain with brain dissections.

Although hr38 and sr mRNA levels peak at 60 min with little or no changes evident at 15 min and

even at 30 min (Figure 2F,G), the nascent RNA data indicate that the enhanced transcription of

both genes is obvious at the first 15 min time point and peaks at 30 min, 30 min before the peak

time of mRNA (Figure 2F,G). For shorter genes such as CG14186 (5.7 kb), the stimulation-induced

increases are smaller at the nascent RNA level, presumably because much less time is required for

transcription elongation and polyadenylation. More importantly and unlike the longer genes, there is

no delay between the nascent and mRNA peak times for the shorter genes (Figure 2H). The data

taken together indicate that the long gene lengths of some prominent ARGs reduce the mRNA

induction rates due to the time required for nascent events like transcriptional elongation.

ARGs induced by two additional stimulation paradigmsTo test whether these ARGs are universal and respond to different stimulation regimens, we

employed two additional paradigms: dTrpA1-mediated firing and high KCl-induced depolarization

(Figure 3A). UAS-dTrpA1 was overexpressed under the control of Elav-GAL4, and flies were stimu-

lated in vivo by shifting the temperature from 18˚C to 29˚C to stimulate the temperature-sensitive

dTrpA1 cation channel. High KCl stimulation was done ex vivo by incubating dissected brains in 90

mM KCl. In this case, the voltage sensor ArcLight was used to confirm successful depolarization

(Video 1, Figure 3—figure supplement 1) (Cao et al., 2013).

Hundreds of genes are induced with the dTrpA1 paradigm but many are also robustly heat-

induced, i.e., they are also induced in control flies expressing only Elav-GAL4. Nonetheless, 97 genes

including hr38 and sr were induced by dTrpA1 at 60 min but not induced or induced much less

robustly in the control flies (Figure 3B, Figure 3—source data 1). With KCl treatment, only 38 genes

were significantly induced in 60 min, and most of them are stress-induced such as heat shock genes

(Figure 3—source data 2). Hr38 and sr were also significantly induced by KCl but with much smaller

increases (2.5 and 1.5 fold, respectively) compared to the other two paradigms (Figure 1—source

data 1, Figure 3—source data 1). To accommodate genes that might be induced more slowly with

the KCl paradigm, an extra 90 min time point was assayed (three biological replicas). Because many

more ARGs (107 genes in total) were significantly induced and had bigger induction amplitudes at

90 min compared to 60 min (Figure 3—source data 3, Figure 3C), this 90 min time point was com-

pared with the other two induction methods.

Only 12 ARGs are shared by all the three paradigms (including hr38 and sr; Figure 3D,E, Fig-

ure 3—source data 4). To verify this conclusion, qPCR was used on 6 of these shared genes, and

Figure 2 continued

baseline, FPKM > 10 with stimulation), and ChR2-XXL-induced ARGs (red). Only the longest isoform of each gene is plotted. (D) Correlation of ChR2-

XXL-induced ARGs gene length (y-axis) and fold changes of 15 min compared to 0 min (abbreviated as FC15/0, x-axis). (E) Correlation of ChR2-XXL-

induced ARGs gene length (y-axis) and fold changes of 60 min compared to 0 min (abbreviated as FC60/0, x-axis). (F–H) Expression profiles of hr38, sr,

and CG14186 from fly heads 0, 15, 30 and 60 min after LED stimulation. Nascent RNA levels are shown in green and mRNA levels in blue. Numbers to

the right of the expression graphs indicate data range.

DOI: 10.7554/eLife.19942.005

The following figure supplement is available for figure 2:

Figure supplement 1. Gene structure of some long ARGs induced by ChR2-XXL, including hr38, sr, CG8910, and CG11221.

DOI: 10.7554/eLife.19942.006

Chen et al. eLife 2016;5:e19942. DOI: 10.7554/eLife.19942 5 of 21

Research article Neuroscience

Figure 3. ARGs induced by two additional stimulation paradigms. (A) Schemes of two additional stimulation paradigms: dTrpA1 and KCl. (B) Heat map

of dTrpA1-induced ARG expression with 0, 15, 30 and 60 min temperature shift (normalized to 0 min expression). N = 3 biological replicates. (C) Heat

map of KCl-inducd ARG expression with 0, 15, 30, 60 and 90 min KCl incubation (normalized to 0 min expression). N = 3 biological replicates. (D)

Overlap of ARGs among the three different stimulation paradigms: ChR2-XXL, dTrpA1 and KCl (90 min). (E) Gene expression of the 12 shared ARGs

induced by all three stimulation paradigms. N = 3 biological replicates, error bars represent ±SEM, **p<0.01, ***p<0.001, ****p<0.0001, exact test. (F)

Gene expression of paradigm-specific ARGs: pk61c and skl are specific to the ChR2-XXL paradigm; CG13054 and grasss are specific to the dTrpA1

paradigm; and CG13322 and CG10383 are specific to the KCl paradigm. N = 3 biological replicates, error bars represent ±SEM, n.s. represents non-

significant, **p<0.01, ***p<0.001, ****p<0.0001, exact test.

DOI: 10.7554/eLife.19942.007

The following source data and figure supplements are available for figure 3:

Source data 1. dTrpA1-induced ARGs in fly brains.

DOI: 10.7554/eLife.19942.008

Source data 2. KCl-induced ARGs in 60 min in fly brains.

Figure 3 continued on next page

Chen et al. eLife 2016;5:e19942. DOI: 10.7554/eLife.19942 6 of 21

Research article Neuroscience

the results of all six are qualitatively consistent with the high-throughput RNA sequencing results

(Figure 3—figure supplement 2A).

The overlap between the three different pairwise combinations is larger: 13 genes are shared

between dTrpA1 (13% of dTrpA1-induced ARGs) and KCl (12% of KCl-induced ARGs), and 24 genes

between ChR2-XXL (25% of ChR2-XXL-induced ARGs) and KCl (22% of KCl-induced ARGs). The larg-

est number (35) and fraction of genes are shared between ChR2-XXL (36% of ChR2-XXL-induced

ARGs) and dTrpA1 (36% of dTrpA1-induced ARGs). These data indicate that many ARGs respond to

firing independent of the stimulation paradigm (Figure 3D, Figure 3—source data 4).

There are however some genes that are specific to the stimulation paradigm, consistent with

what has been observed in mammals (Bepari et al., 2012; Fields et al., 1997). For example, pk61c

and skl are induced only by ChR2-XXL; CG13054, whereas grass is induced only by dTrpA1;

CG13322 and CG10383 are only induced by KCl (Figure 3F). These stimulation paradigm-specific

ARGs and six others were verified by qPCR (Figure 3—figure supplement 2B), with conclusions

qualitatively consistent with high-throughput RNA sequencing results.

Stronger stimulation paradigms, for example

longer LED exposure for ChR2-XXL flies, bigger

temperature shift for dTrpA1 flies or incubation

with higher KCl concentration, may further

increase the overlap among paradigms. To test

this, the LED exposure of the ChR2-XXL flies was

extended from 30 s to 30 min. 2 out of 8 para-

digm-specific ARGs that were negative with the

30 s exposure were positive with the 30 min stim-

ulation, but the other six remained negative (Fig-

ure 3—figure supplement 3). In addition, we

tested by qPCR whether some of the KCl-specific

ARGs are due to the ex vivo condition only used

for the KCl-stimulation: 4 out of 4 tested are not

induced by the other two stimulation paradigms

performed under the same ex vivo condition, i.e.,

exposure of dissected brains to the LED or to

temperature shift (Figure 3—figure supplement

4A). In contrast, four common ARGs were stimu-

lated ex vivo by ChR2-XXL and dTrpA1 in 60 min

to levels quite comparable to those from the KCl

paradigm at 90 min, suggesting that the different

results are due to the different stimulation para-

digms rather than ex vivo vs. in vivo stimulation

(Figure 3—figure supplement 4B).

Figure 3 continued

DOI: 10.7554/eLife.19942.009

Source data 3. KCl-induced ARGs in 90 min in fly brains.

DOI: 10.7554/eLife.19942.010

Source data 4. Overlapped ARGs in brains.

DOI: 10.7554/eLife.19942.011

Figure supplement 1. Fluorescence changes of ArcLight indicate successful depolarization of neurons with KCl stimulation.

DOI: 10.7554/eLife.19942.012

Figure supplement 2. qPCR verification of high-throughput RNA sequencing results.

DOI: 10.7554/eLife.19942.013

Figure supplement 3. qPCR quantification of paradigm-specific ARGs in ChR2-XXL flies with a 30 min LED stimulation followed by another 30 min

recovery.

DOI: 10.7554/eLife.19942.014

Figure supplement 4. qPCR quantification of ARGs induction under ex vivo condition.

DOI: 10.7554/eLife.19942.015

Video 1. S1 Live imaging of Arclight in PDF+ neurons

in response to KCl treatment. To show KCl treatment

can successfully depolarize neurons, Arclight was

artificially expressed in PDF+ neurons (Pdf-GAL4;UAS-

ArcLight). Arclight signal was monitored with

fluorescent microscope while KCl was perfused into the

chamber containing a dissected brain. Quantification of

fluorescent intensity is shown in Figure 3—figure

supplement 1.

DOI: 10.7554/eLife.19942.016

Chen et al. eLife 2016;5:e19942. DOI: 10.7554/eLife.19942 7 of 21

Research article Neuroscience

The results taken together suggest that most paradigm-specific ARGs will remain responsive to

only one type of stimulation despite the fact that stronger stimulation will somewhat increase the

overlap among the three paradigms, e.g., the distinction between the paradigms might be quantita-

tive as well as qualitative. In any case, the data taken together significantly extend the list of bona

fide ARGs in flies.

Transcriptome profiling in DA and PDF+ neuronsFly brains contain many types of neurons with distinct properties as well as glia and other tissues

(The Interactive Fly, http://www.sdbonline.org/). Transcript changes in specific neurons may there-

fore be invisible with assays from brains and heads, especially if there is significant cell-type specific-

ity in the transcriptional response to neuronal firing. To address this possibility, we performed

mRNA deep sequencing on sorted green fluorescent protein (GFP) labeled DA and PDF+ neurons

expressing dTrpA1 with or without a temperature shift, namely, the same dTrpA1 paradigm used for

brains. DA and PDF+ neurons are two distinct populations and serve different functions in flies. Each

brain has about ~120 DA neurons and ~16 PDF+ neurons. Flies expressing only the GAL4 drivers

were used as a control for heat-induced genes.

Despite much less starting material than brains, expression profiles from both cell types are well

correlated between replicates (R2 > 0.9 between replicates) (Figure 4A,B). In contrast, baseline

expression profiles among cells and tissues (brains, PDF+ and DA neurons) are distinct as expected

(Figure 4—figure supplement 1). Previously identified cell-type enriched transcripts were used as

markers to ensure that the correct neurons were collected. The housekeeping gene Act5C is

expressed ubiquitously in brains, DA and PDF+ neurons (Figure 4C). The DA neuron-enriched tran-

script ple (encodes tyrosine hydroxylase) and the PDF+ neuron-specific transcript pdf are both highly

Figure 4. Transcriptome profiling in sorted DA and PDF+ neurons. (A) Gene expression (log10 FPKM) plot among three biological replicates (baseline

levels) in DA neurons. Only genes with FPKM > 0.01 are plotted. (B) Gene expression (log10 FPKM) plot among three biological replicates (baseline

levels) in PDF+ neurons. Only genes with FPKM > 0.01 are plotted. (C) Expression levels of the house-keeping gene (act5c) and marker genes (ple

specific for DA neurons and pdf specific for PDF+ neurons) in all three tissues. Numbers to the right of the expression graphs represent data range.

DOI: 10.7554/eLife.19942.017

The following figure supplement is available for figure 4:

Figure supplement 1. Transcriptomes of brains, DA and PDF+ neurons are very distinct.

DOI: 10.7554/eLife.19942.018

Chen et al. eLife 2016;5:e19942. DOI: 10.7554/eLife.19942 8 of 21

Research article Neuroscience

enriched only in the corresponding neuron types, indicating the correct cell types were collected

(Figure 4C) (Budnik and White, 1987; Helfrich-Forster, 1997).

There are many neuron-specific ARGs85 candidate ARGs in DA neurons and 81 in PDF+ neurons are significantly induced in dTrpA1 flies

but not in control flies (Figure 5A, Figure 5—source data 1 and 2). There was also very limited

overlap between the three sets of ARGs: only six genes are in common between all three sources,

brains, DA and PDF+ neurons (Figure 5A,B); 11 additional genes are shared between brains and DA

Figure 5. Many ARGs are tissue-specifically regulated in response to firing. (A) Overlap of ARGs in different tissues. (B) Gene expression of shared

ARGs induced in all tissue types. N = 3 biological replicates, error bars represent ±SEM, *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001, exact test. (C)

Gene expression of tissue-specific ARGs. N = 3 biological replicates, error bars represent ±SEM, n.s. represents non-significant, *p<0.05, **p<0.01,

***p<0.001, ****p<0.0001, exact test.

DOI: 10.7554/eLife.19942.019

The following source data and figure supplements are available for figure 5:

Source data 1. ARGs induced in DA neurons.

DOI: 10.7554/eLife.19942.020

Source data 2. ARGs induced in PDF+ neurons.

DOI: 10.7554/eLife.19942.021

Source data 3. Overlapped ARGs in different tissue types.

DOI: 10.7554/eLife.19942.022

Figure supplement 1. Not all common ARGs in brains are induced in DA and PDF+ neurons.

DOI: 10.7554/eLife.19942.023

Figure supplement 2. Big overlap among heat-induced genes in different tissues.

DOI: 10.7554/eLife.19942.024

Chen et al. eLife 2016;5:e19942. DOI: 10.7554/eLife.19942 9 of 21

Research article Neuroscience

neurons; seven between DA and PDF+; and six between brains and PDF+ neurons (Figure 5A, Fig-

ure 5—source data 3). Notably, half of the six common genes are among the 12 genes induced by

all three stimulation paradigms in brains (CG14186, CG17778 and l(1)G0148, Figure 3E), suggesting

that these genes respond to firing in many and perhaps most fly brain neurons (Figure 3—source

data 4). However, the other nine stimulation-independent brain ARGs are not universally induced,

further indicating considerable cell type-specificity of firing-induced gene expression (Figure 5—fig-

ure supplement 1). Previous work on hr38 using in situ hybridization is consistent with this interpre-

tation (Fujita et al., 2013).

Indeed, these data indicate that most ARGs are cell type-specific (Figure 5A). For example,

CG13054 and grass are only significantly induced in brains and DA but not PDF+ neurons; Mtk and

CG8407 are only significantly induced in DA neurons; CG9313 and DAT are exclusively up-regulated

in PDF+ neurons (Figure 5C). There are also genes like JhI-21 and CG10863, which show changes in

sorted DA and PDF+ neurons but smaller or no changes in brains; this suggests masking/dilution

effects in brains (Figure 5C). In contrast, almost complete overlap was observed for heat-induced

genes (Figure 5—figure supplement 2). In sum, a preponderance of cell-type specific ARGs sug-

gests that they endow their neurons with specific functions and lead to different firing-induced

consequences.

Promoter regions of ARGs are at permissive state prior to stimulationChromatin structure plays an important role in regulating gene expression, and promoters (TSS) of

mammalian IEGs are usually located in open chromatin (Fowler et al., 2011). To address the chro-

matin states of fly ARG TSSs, we performed ATAC-seq in fly brains expressing Elav-GAL4; UAS-

dTrpA1 or its sibling control strain expressing Elav-GAL4; CyO before or 30 min after a temperature

shift.

The TSS and its surrounding region (±250 bp) of all annotated fly genes are more accessible than

other regions in the genome as expected (Figure 6—figure supplement 1). This region of ARGs

also shows enhanced accessibility but does not detectably change with firing despite the increase in

transcription (Figure 6A). Even the TSS accessibility of top-ranked ARGs like hr38 (rank #1) and

CG7995 (rank #6) shows no differences with firing (Figure 6B, Figure 6—figure supplement 2). This

is in contrast to the TSS of heat shock protein genes (HSPs), which possesses less accessible chroma-

tin than that of ARGs at baseline but opens substantially by 30 min in both the dTrpA1 and control

flies in response to heat (Figure 6C). In fact, the whole gene body of heat shock genes opens with

heat, consistent with what was shown previously at the hsp70 locus (Petesch and Lis, 2008).

Despite the lack of a firing-induced change in ARG chromatin TSSs (Figure 6A), they are notably

more accessible prior to stimulation compared to all expressed, annotated genes (Figure 6D). Since

baseline ATAC-seq profiles are robust and minimally affected by genetic background (unpublished

data from Rosbash lab), baseline chromatin profiles from ARGs induced by the other paradigms

were pooled for analysis. These ARGs, like the dTrpA1-induced ARGs, also show higher accessibility

than the rest of the genome (Figure 6D). This is unlikely due to higher baseline expression levels;

only KCl-induced ARGs show significantly higher baseline expression levels despite showing interme-

diate chromatin accessibility of the three firing paradigms (Figure 6—figure supplements

3, Figure 6D). In conclusion, ARG TSS regions are generally more accessible at baseline, presumably

to allow rapid transcriptional induction upon stimulation.

Generation of luciferase reporters for neuronal activitySince little was previously known about fly ARGs, no useful neuronal activity reporters comparable

to c-fos have been constructed. We therefore used the information in this study to generate several

luciferase reporters for monitoring neuronal activity in living flies. We used the online motif enrich-

ment tool (http://veda.cs.uiuc.edu/cgi-bin/MET/interface.pl) to identify transcription factor binding

sites in either the 1 kb or 5 kb upstream region of ChR2-XXL-induced ARGs; expressed genes

(FPKM > 0 at baseline and FPKM > 10 with stimulation) were used as a null set for background sub-

traction. Lola (significance = 4.13E-07), eip78C (significance = 5.92E-05), relish (rel,

significance = 1.78E-04), broad (br, significance = 6.62E-02, insignificant) and cf2 (significance = 0.43,

insignificant) were chosen to represent different levels of enrichment ranking, from high to low (Fig-

ure 7—source data 1 and 2). Three binding sites of each transcription factor were multimerized and

Chen et al. eLife 2016;5:e19942. DOI: 10.7554/eLife.19942 10 of 21

Research article Neuroscience

inserted into a vector containing mCherry and luciferase, which originated from the Yin lab

(Figure 7A) (Tanenhaus et al., 2012), and transgenic flies generated.

To assay the reporters, they were crossed to Elav-GAL4;UAS-flp;UAS-ChR2-XXL. In progeny

expressing all the transgenes, GAL4 is expressed pan-neuronally and activates expression of ChR2-

XXL to fire neurons. GAL4 also activates flp, which removes sequences between the two FRT sites in

the reporter; this allows brain-specific expression of luciferase under the control of the enhancer, the

transcription factor-multimer, and the promoter (Figure 7A). Adult progeny were subjected to a pro-

tocol like in Figure 1: adult flies were subjected to a single 30 s blue LED exposure, after which their

luciferase levels were monitored every 10–15 min for several hours with a top counter

(Brandes et al., 1996). All flies expressing ChR2-XXL showed seizure behavior within seconds of

LED exposure as described above, whereas the control flies without ChR2-XXL expression were

unaffected.

Figure 6. Promoter regions of ARGs are at permissive state prior to stimulation. (A) Average chromatin accessibility at TSS of dTrpA1-induced ARGs in

brains expressing Elav-GAL4;UAS-dTrpA1 or Elav-GAL4/CyO (control) at 0 or 30 min of temperature shift. TSS is aligned to 0 at x-axis which indicates

the distance from TSS in base pair. Negative numbers indicate upstream of TSS and positive indicates downstream. Same for all the graphs in this

figure. (B) Average chromatin accessibility at TSS of hr38 in brains expressing Elav-GAL4;UAS-dTrpA1 or Elav-GAL4/CyO (control) at 0 or 30 min of

temperature shift. (C) Average chromatin accessibility at TSS of heat-shock proteins in brains expressing Elav-GAL4;UAS-dTrpA1 or Elav-GAL4/CyO

(control) at 0 or 30 min of temperature shift. (D) Average chromatin accessibility at TSS of all expressed genes (FPKM > 0.1) and ARGs induced with

different stimulation paradigms. Data are the average of all different conditions (genotypes and time points) and no changes of chromatin accessibility

were observed among these conditions.

DOI: 10.7554/eLife.19942.025

The following figure supplements are available for figure 6:

Figure supplement 1. Promoter regions of all annotated genes are more accessible in brains.

DOI: 10.7554/eLife.19942.026

Figure supplement 2. Chromatin accessibility of TSS of CG7995 show no changes with firing.

DOI: 10.7554/eLife.19942.027

Figure supplement 3. Expression levels of ARGs at baseline levels.

DOI: 10.7554/eLife.19942.028

Chen et al. eLife 2016;5:e19942. DOI: 10.7554/eLife.19942 11 of 21

Research article Neuroscience

The three transcription factors (LOLA, EIP78C, and REL) with high significance showed a dramatic

increase in luciferase signal immediately after LED exposure, and the signal decayed in about an

hour (Figure 7B). In contrast, the two transcription factors with much lower significance (br, cf2)

showed no change (Figure 7—figure supplement 1.). Although the luciferase levels are somewhat

noisy, they show that the three transcription factors with high significance are rapidly activated in

response to firing. We also note that the different reporters may be activated in different neurons, i.

Figure 7. Generation of luciferase reporters for in vivo monitoring of neuronal activity. (A) Schemes of transgenes constructed for monitoring neuronal

activity (top) and other transgenes crossed to be expressed in the same flies. Transcription factor binding sites include binding sites of LOLA, EIP78C

and REL. Three of each transcription factor binding site were multimerized and fused upstream to the promoter and individually injected to fly

embryos. These reporter-containing flies were then crossed to the existing Elav-GAL4, UAS-FLP and UAS-ChR-XXL (not in controls) flies. (B) Luciferase

levels of flies expressing the constructed reporters exposed to a 30 s LED stimulation recorded every 10–15 min for ~4 hr. Gray indicates luciferase

levels of control flies expressing all the transgenes in (A) except for UAS-ChR2-XXL. Red indicates flies expressing all the transgenes in (A). The

experimental and control flies were put in the same 96-well plate and stimulated and recorded in parallel. The 30 s LED exposure is indicated by the

black bars below the graphs. N = 8 flies. Representative images of multiple biological replicates are shown. Shaded regions represent ±SD.

DOI: 10.7554/eLife.19942.029

The following source data and figure supplement are available for figure 7:

Source data 1. Motif enrichment in the 5 kb upstream regions of ChR2-XXL-induced ARGs.

DOI: 10.7554/eLife.19942.030

Source data 2. Motif enrichment in the 1 kb upstream regions of ChR2-XXL-induced ARGs.

DOI: 10.7554/eLife.19942.031

Figure supplement 1. Luciferase reporters for br and cf2 show no change with LED stimulation in both experiment (red) and control (gray) flies.

DOI: 10.7554/eLife.19942.032

Chen et al. eLife 2016;5:e19942. DOI: 10.7554/eLife.19942 12 of 21

Research article Neuroscience

e., the results do not imply that we have identified a fos-eqivalent for Drosophila. However, they do

indicate that the strategy is sound and should work for other activity-activated transcription factors

and for the design of additional reporters, including for cell-specific neuronal activity (Guo et al.,

2016).

DiscussionUsing high-throughput sequencing, we identified ARGs in fly brains in response to ChR2-XXL,

dTrpA1 and KCl-induced neural firing. Each stimulation paradigm induced expression of about 100

genes, and there is substantial overlap between the 3 sets of ARGs (Figure 3D). The most robust

ARGs are the two known genes identified previously, hr38 and sr; they are induced dramatically by

ChR2-XXL and dTrpA1 (Figure 3E). They are also induced with KCl albeit with smaller amplitudes

(Figure 3E). 10 additional genes are also induced by all three protocols, and even more genes are

induced by at least two paradigms (Figure 3D). Although the results dramatically increase the num-

ber of common ARGs, there are also a large number of specialized ARG candidate genes

(Figure 3D). Not surprisingly perhaps, KCl has the most different ARGs. This is presumably because

it is the only in vitro stimulation paradigm and stimulates the whole brain including non-neuronal

cells and tissues.

We also used the dTrpA1 paradigm to identify ARGs in sorted neurons. Many of them are specific

to neuron type, and even the most robust brain ARGs are not necessarily induced by firing in every

region of the brain (Figure 5, Figure 5—figure supplement 1). This is in contrast to heat-induced

genes, most of which are shared by the three sources of RNA (Figure 5—figure supplement 2.).

Heat-induced gene expression unfortunately creates a complicated situation: some putative spe-

cific ARGs are also induced with the 18˚C to 29˚C shift in control (no dTrpA1) samples. These genes

are thus excluded from the ARG list because they are defined as heat-induced by the cutoffs used in

the analysis. Because it is possible that the temperature shift can lead to neuronal firing, e.g., by

activating endogenous dTrpA1, there is currently no clear distinction between heat- and firing-

induced genes. Future experiments with other tools like ChR2-XXL may be able to provide more

information. In any case, the neuron-specific ARGs probably contribute to different physiological

responses, which contribute in turn to distinct behavioral responses (Guo et al., 2014; Ueno et al.,

2012). The results also provide good candidates as neural activity markers, e.g., Figure 7, and sug-

gest that multiple genes should be tried, to accommodate cell type as well as firing heterogeneity.

Although the in vivo nature of our most important stimulation paradigms precluded using inhibi-

tors to assay the dependence of ARGs on de novo protein synthesis (data not shown), the induction

time scales of these fly ARGs, 60–90 min, resemble those of mammalian IEGs. A comparison

between fly ARGs with mammalian IEGs therefore seems appropriate.

Whereas mammalian IEGs are highly enriched in transcription factors (Spiegel et al., 2014), they

are less enriched in the fly brain and sorted neuron ARGs: ChR2-XXL induced ARGs contain only

seven transcription factors among 96 genes; this is a much smaller fraction than in mammals

(Figure 1D). KCl-induced ARGs show no enrichment in functions, and dTrpA1-induced ARGs in the

cell types have a weak enrichment in different functions, such as steroid hormone receptor activity in

brains, imaginal disc growth factor receptor binding in DA neurons and organic anion transmem-

brane transporter activity in PDF+ neurons (Table 1). The weak enrichment suggests diverse ARG

functions, which resemble more closely mammalian DRGs or SRGs rather than PRGs (see below).

Less transcription factor activity and more diverse functions are consistent with the fact that fly

homologues of the two most universal mammalian IEGs c-fos and c-jun, kayak and jra respectively,

are not induced or poorly induced under most conditions. This is especially true for kayak; jra is

Table 1. Gene ontology analysis on dTrpA1-induced ARGs in different tissues with GOrilla.

Functions p-value Genes

Brains Steroid hormone receptor activity 0.000536 3

DA Imaginal disc growth factor receptor binding 0.000843 4

PDF+ Organic anion transmembrane transporter activity 0.000742 5

DOI: 10.7554/eLife.19942.033

Chen et al. eLife 2016;5:e19942. DOI: 10.7554/eLife.19942 13 of 21

Research article Neuroscience

induced by both dTrpA1 and KCl in brain but with quite small fold changes (Figure 3—source data

1 and 2).

We also noticed that most fly ARGs show rather small fold-changes upon stimulation (Figure 8).

It is unlikely that this is due to brain tissue heterogeneity as similar results were obtained with the

sorted neurons. A similar weak induction of seizure-induced genes was previously reported in fly

heads (Guan et al., 2005). As only the first 60–90 min after stimulation was assayed, it is possible

that the small fold-changes reflects a slower induction time course in flies than in mammals. Another

possibility is that the nature of the firing, e.g., firing rate, has an impact on the induction amplitudes,

time courses and even the genes induced. The differences between the three stimulation paradigms

(Figure 3) indicate that this possibility may have some merit. Nonetheless and taken together with

the stimulation-paradigm and cell type-specific induction of ARGs, the modest fold-changes at 60–

90 min helps explain the long-standing failure to identify bona fide and well-accepted ARGs in flies.

Instead of having shorter average gene lengths like in mammals, fly ARGs (brain as well as sorted

neurons) show even longer gene length distributions than average fly genes (Figure 2C, Figure 9).

Several robust ARGs are above 10 kb with large first introns (Figure 2—figure supplement 1). For

example, hr38 is ~31 kb and shares high protein sequence identity with the mammalian protein

encoded by Nr4a2; it spans only ~16.8 kb. sr, the shorter isoform of which is induced by firing,

spans ~11.7 kb, whereas EGR3, the mouse gene that encodes a protein with high identity to SR, is

only ~5.2 kb in length. Considering the much smaller genome size and gene size of flies compared

to mammals and the potential slower transcription elongation rates in flies, the long gene lengths

and large first introns of fly ARGs may counterintuitively serve to significantly slow down induction

rates (Ardehali and Lis, 2009). Another possibility is that the large gene size reflects important regu-

latory elements for firing-mediated transcription embedded in the introns of these genes. Perhaps

they are relevant to the diversity and cell type specificity of most fly ARGs.

From this point of view, fly ARGs resemble more closely mammalian DRGs or even SRGs than

IEGs (Spiegel et al., 2014; Tullai et al., 2007). Mammalian IEGs are largely shared between different

types of neurons such as excitatory and inhibitory neurons (Spiegel et al., 2014), but fly ARGs like

Figure 8. Fold changes distribution of all ARGs. X-axis indicates fold changes (log2) at 60 or 90 (for KCl) min. Y-axis indicates the frequency of genes.

ChR2-XXL indicates ChR2-XXL-induced ARGs. dTrpA1 (brains) indicates dTrpA1-induced genes in brains. KCl (60 min/90 min) indicates KCl-induced

ARGs in 60 or 90 min. DA indicates dTrpA1-induced ARGs in DA neurons. PDF+ indicates dTrpA1-induced ARGs in PDF+ neurons.

DOI: 10.7554/eLife.19942.034

Chen et al. eLife 2016;5:e19942. DOI: 10.7554/eLife.19942 14 of 21

Research article Neuroscience

SRGs appear quite tissue-type specific, i.e., many genes are only induced in one of the tissue types

tested despite similar baseline levels (Figure 5) (Mardinly et al., 2016). This is in line with the fact

that fly neurons appear quite heterogeneous, e.g., they have distinct transcript enrichment and

electrophysiological properties (Liang et al., 2016; Nagoshi et al., 2010). Although unlikely in our

view, it remains possible that PDF+ and DA neurons are outliers and that ARGs in most other neu-

rons are less cell type-specific. Nonetheless, we hypothesize that specific ARGs are an important fea-

ture of neuronal identity and play a role in neuron-specific physiological properties. As an example,

preliminary analysis indicates that the PDF cell-ARGs are a subset of the mRNAs that undergo circa-

dian oscillations within these neurons ([Kula-Eversole et al., 2010] and data not shown).

The non-identical ARGs with the different paradigms also suggest different mechanisms for

responding to different modes of stimulation. A less stereotyped response may be generally charac-

teristic of invertebrates. Even non-mammalian vertebrates may be quantitatively if not qualitatively

different from mammals. For example, induction of IEGs like fos and egr1 was also observed in birds

and zebrafish but with smaller increases (about 4 to 6-fold) than human and mice (usually well above

10-fold) (Bebien et al., 2003; Burmeister and Fernald, 2005; Lopes et al., 2015; Mokin and Kei-

fer, 2005; Moore and Whitmore, 2014). However, more high-throughput data are required to

determine if the non-mammalian vertebrates resemble more closely flies or mammals. In addition,

different stimulation paradigms were employed in the non-mammalian vertebrate studies, making

the data less comparable. In any case, one can imagine that mammals need a more potent initial

transcriptional system than flies to respond to firing. Not only are mammalian neurons bigger with

more and longer processes, but they also have much larger genomes (longer genomic DNA and

more enhancers to bind); both of these factors may require higher levels of transcripts to generate a

required higher abundance of transcription factors. The increase may be required to activate the

thousands of enhancers in mammals rather than the hundreds in flies, and/or more structural pro-

teins are needed to change the morphology of ‘giant’ mammalian neurons.

Nonetheless, one feature shared by fly ARGs and mammalian IEGs is the relatively permissive

state of the TSS chromatin structure (Fowler et al., 2011). ATAC-seq shows that the chromatin

accessibility of the ARG TSS is higher than that of other genes (Figure 6D), and chromatin architec-

ture contributes to gene expression regulation (Wu, 1997). Although no further opening is observed

with firing, it is possible that more subtle changes such as histone modifications occur and are not

Figure 9. Gene length distribution of all annotated genes and ARGs. All indicates all annotated genes. Expressed indicates genes with FPKM > 0 at

baseline, FPKM > 10 with stimulation in brains. Brains indicate dTrpA1-induced ARGs in brains. DA indicates dTrpA1-induced ARGs in DA neurons. PDF

+ indicates dTrpA1-induced ARGs in PDF+ neurons. KCl indicates KCl-induced ARGs in 90 min.

DOI: 10.7554/eLife.19942.035

Chen et al. eLife 2016;5:e19942. DOI: 10.7554/eLife.19942 15 of 21

Research article Neuroscience

detectable with ATAC-seq (Figure 6A). In the cases of mammals, histones around IEGs promoters

are modified to be more accessible to transcription factors at baseline; the modification levels could

be even higher with firing (Kim et al., 2010). In any case, the data taken together suggest some

blurring of the distinction between IEGs and SRGs in fly ARGs: open chromatin and rapid induction

but a diverse set of neuron-specific functions.

Finally, we successfully generated luciferase reporters for in vivo neural activity monitoring based

on the ChR2-XXL data (Figure 7). Rather than using the regulatory region of a single gene to con-

struct a reporter, we generalized a previous strategy (Tanenhaus et al., 2012) and used enriched

transcription factor binding sites combined with FRT sites for cell type specific expression and a

short-lived luciferase gene. This was not only to amplify the signal but also to avoid the situation in

which a particular promoter is only responsive in certain neurons or to certain types of stimulation.

We thought this was likely considering that ARGs are stimulation paradigm- and cell type-specific.

Indeed, we tried many Janelia Research Campus GAL4 drivers, which use different regions of the

hr38, sr and cbt promoters for neural activity monitoring, but they were all either unresponsive or

generated very low luciferase levels (data not shown).

Two of the three positive reporters (Figure 7) work well for circadian monitoring of neuronal

activity in PDF+ neurons (data not shown), suggesting the signal to noise ratio is sufficient for moni-

toring small subsets of neurons. This result and others indicate that our strategy as well as these spe-

cific reporters can provide the temporal and spatial resolution for in vivo monitoring of small

numbers of specific neurons. The strategy also builds a foundation for future elaboration and

optimization.

Materials and methods

Drosophila stocksDrosophila (RRID:FlyBase_FBst1021211) were reared on standard cornmeal/agar medium supple-

mented with yeast under 12 hr LD cycles at 25˚C. Elav-GAL4 (C155) was obtained from Bloomington

Stock Center. UAS-ChR2-XXL was a kind gift from Kittel Lab. UAS-dTrpA1was from Dr. Paul Garrity.

TH-GAL4 was described in (Guo et al., 2014). Pdf-GAL4 was described in (Stoleru et al., 2004).

UAS-mCD8GFP;Pdf-GAL4 was described in (Nagoshi et al., 2010). Flies expressing both GAL4 and

UAS-dTrpA1 were reared at 18˚C until tested. UAS-FLP (#55804) was obtained from Bloomington

Stock Center.

Stimulation protocolsFor the ChR2-XXL paradigm, both the control and experimental flies were transferred from normal

fly food to all trans-retinal (ATR) – containing food for one day and kept in dark until stimulation.

ATR powder was dissolved in EtOH into 100 mM stock, and further diluted to 0.4 mM as final work-

ing concentration (Klapoetke et al., 2014). To fire neurons, 30 s 10 Hz 0.7 V blue LED with 5 ms

pulse width was given to flies. Flies were allowed to recover in dark for 15, 30, 60 min.

For the dTrpA1 paradigm, both the control and experimental flies were reared at 18˚C and

shifted to 29˚C for 0, 15, 30, 60 min to fire neurons.

For the KCl paradigm, Canton-S WT fly brains were dissected in adult hemolymph-like medium

(AHL) consisting 108 mM NaCl, 5 mM KCl, 2 mM CaCl2, 8.2 mM MgCl2, 4 mM NaHCO3, 1 mM

NaH2PO4, 5 mM trehalose, 10 mM sucrose and 5 mM HEPES and transferred to depolarization

buffer containing 28 mM NaCl, 85 mM KCl, 2 mM CaCl2, 8.2 mM MgCl2, 4 mM NaHCO3, 1 mM

NaH2PO4, 5 mM trehalose, 10 mM sucrose and 5 mM HEPES (Shang et al., 2013; Wang et al.,

2003). pH for both buffers was adjusted to 7.5, and mOsm to 265.

For ChR2-XXL and dTrpA1 stimulation performed under ex vivo condition, fly brains were dis-

sected in AHL and subjected to either a 30 s LED exposure or a 60 min 29˚C water bath incubation.

For all experiments assayed from brains and heads, 8–10 brains (50% males and females) from ~7

day-old young adults were used for each condition. Brains were carefully dissected to remove all

non-brain tissues, and only intact undamaged brains were used for RNA extraction. For sorted neu-

rons, 20 young males and 20 young females were used.

Chen et al. eLife 2016;5:e19942. DOI: 10.7554/eLife.19942 16 of 21

Research article Neuroscience

ArcLight imagingFly brains were dissected in AHL and then mobilized using a pin anchored to the chamber bottom

laid with Sylgard (Dow Corning, Midland, MI) (Shang et al., 2013; Wang et al., 2003). Depolariza-

tion buffer was perfused into the chamber using a gravity-fed ValveLink perfusion system (Automate

Scientific, Berkeley, CA) (Haynes et al., 2015; Shang et al., 2013). Imaging was performed with an

Olympus BX51WI fluorescence microscope (Olympus, Center Valley, PA) under an Olympus x60

(0.90W, LUMPlanFI) water-immersion objective, and captured using a charge-coupled device camera

(Hamamatsu ORCA C472-80-12AG). The following filter sets were used for excitation and emission

(Chroma Technology, Bellows Falls, VT): excitation, HQ470/x40; dichroic, Q495LP; emission, HQ525/

50m. mManager was used for recording with 2 Hz with 4 x binning with 500 ms exposure time and

50 ms intervals (Edelstein et al., 2010).

RNA extraction, amplification for sequencing libraries and qPCRvalidationFly brains or heads were isolated for RNA extraction using TRIzol. mRNA libraries were constructed

following Illumina Trueseq protocol. Nascent RNA libraries were constructed as described in

(Rodriguez et al., 2013). mRNA libraries from sorted neurons were prepared as described in

(Abruzzi et al., 2015). ATAC-seq libraries were done using Nextera DNA Sample Prep Kit. Libraries

were sequenced using HiSeq, NextSeq or MiSeq sequencers.

For qPCR validation, total RNA was reverse-transcribed with random primer (Promega, C1181)

and gene expression was then quantified with SYBR Green PCR Master Mix using primers listed in

Supplementary file 1.

Sequencing data analysisRNA libraries were mapped to Drosophila genome dm3 using Tophat (Trapnell et al., 2009) with

default setting (e.g. two mismatches allowed) and expression levels were quantified using Cufflinks

(Trapnell et al., 2012). FPKM > 0 for un-stimulated groups and FPKM > 10 for stimulated groups

were used as expression cutoff. HTSeq (Anders et al., 2015) was used to produce raw read count

for each genes. EdgeR (McCarthy et al., 2012; Robinson et al., 2010; Robinson and Smyth, 2007,

2008; Zhou et al., 2014) was used for statistical analysis. N = 3 biological replicates for libraries

from brains. N = 3–4 biological replicates for libraries from sorted neurons. p-value < 0.01 was

defined as significance except for PDF+ ARGs (p-value < 0.05 due to slightly bigger variation). To

distinguish heat- or firing-induced ARGs in dTrpA1 paradigms, t-test was used among replicates

between the experiment and control flies. ARGs are defined to be firing-induced only when

P-value < 0.05 and experiment flies show bigger induction than control flies. For ATAC-seq, Trim

Galore (http://www.bioinformatics.babraham.ac.uk/projects/trim_galore/) is used to trim low quality

bases (quality score < 20) on either end of the reads and remove adapter/tagmentation sequences.

Bowtie2 (Langmead and Salzberg, 2012) was used to map the reads to reference genome with

parameters ‘-D 20 -R 3 -N 1 -L 20’. Alignments below a score of 10 were discarded, and PCR dupli-

cates were removed using Picard (version 1.119). Chromatin accessibility at TSS were pooled and

averaged as described in (Menet et al., 2014).

Generation of luciferase reporter fliesThe cre-luc vector was a kind gift from Jerry Yin lab. The cre sites of the vector were swapped with

binding site multimers (3X) of transcription factors including LOLA, EIP78C, REL, BR and CF2 using

the primers in Supplementary file 1. Transgenes were injected into w1118 flies with the P-element

transformation service of BestGene Inc.

AcknowledgementsWe thank Drs. Oliver Hobert and Haosheng Sun from Columbia University for comments on an early

version of the manuscript. We also thank Drs. Suzanne Paradis, Michael Marr, Nelson Lau, Patrick

Emery and Leslie Griffith, for comments on a near-final version. Many thanks go to members of Ros-

bash lab for thoughtful discussion, and especially Kate Abruzzi for comments.

Chen et al. eLife 2016;5:e19942. DOI: 10.7554/eLife.19942 17 of 21

Research article Neuroscience

Additional information

Funding

Funder Author

Howard Hughes Medical Insti-tute

Xiao ChenReazur RahmanFang GuoMichael Rosbash

The funders had no role in study design, data collection and interpretation, or thedecision to submit the work for publication.

Author contributions

XC, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or

revising the article; RR, Analysis and interpretation of data, Drafting or revising the article; FG, Pro-

vided advice and help, Conception and design; MR, Conception and design, Analysis and interpreta-

tion of data, Drafting or revising the article

Author ORCIDs

Michael Rosbash, http://orcid.org/0000-0003-3366-1780

Additional files

Supplementary files. Supplementary file 1. Supplementary tables. (A) Primers for qPCR validation. (B) Primers for

reporter generation

DOI: 10.7554/eLife.19942.036

Major datasets

The following dataset was generated:

Author(s) Year Dataset title Dataset URL

Database, license,and accessibilityinformation

Xiao Chen 2016 Genome-wide Identification ofNeuronal Activity-regulated Genesin Drosophila

https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE83976

Publicly available atNCBI GeneExpression Omnibus(accession no:GSE83976)

ReferencesAbruzzi K, Chen X, Nagoshi E, Zadina A, Rosbash M. 2015. RNA-seq profiling of small numbers of drosophilaneurons. Methods in enzymology 551:369–386. doi: 10.1016/bs.mie.2014.10.025, PMID: 25662465

Anders S, Pyl PT, Huber W. 2015. HTSeq–a Python framework to work with high-throughput sequencing data.Bioinformatics 31:166–169. doi: 10.1093/bioinformatics/btu638, PMID: 25260700

Ardehali MB, Lis JT. 2009. Tracking rates of transcription and splicing in vivo. Nature Structural Biol Biology 16:1123–1124. doi: 10.1038/nsmb1109-1123, PMID: 19888309

Bebien M, Salinas S, Becamel C, Richard V, Linares L, Hipskind RA. 2003. Immediate-early gene induction by thestresses anisomycin and arsenite in human osteosarcoma cells involves MAPK cascade signaling to Elk-1, CREBand SRF. Oncogene 22:1836–1847. doi: 10.1038/sj.onc.1206334, PMID: 12660819

Bepari AK, Sano H, Tamamaki N, Nambu A, Tanaka KF, Takebayashi H. 2012. Identification of optogeneticallyactivated striatal medium spiny neurons by Npas4 expression. PLoS One 7:e52783. doi: 10.1371/journal.pone.0052783, PMID: 23300775

Bloodgood BL, Sharma N, Browne HA, Trepman AZ, Greenberg ME. 2013. The activity-dependent transcriptionfactor NPAS4 regulates domain-specific inhibition. Nature 503:121–125. doi: 10.1038/nature12743,PMID: 24201284

Brandes C, Plautz JD, Stanewsky R, Jamison CF, Straume M, Wood KV, Kay SA, Hall JC. 1996. Novel features ofdrosophila period transcription revealed by real-time luciferase reporting. Neuron 16:687–692. doi: 10.1016/S0896-6273(00)80088-4, PMID: 8607986

Chen et al. eLife 2016;5:e19942. DOI: 10.7554/eLife.19942 18 of 21

Research article Neuroscience

Budnik V, White K. 1987. Genetic dissection of dopamine and serotonin synthesis in the nervous system ofdrosophila melanogaster. Journal of Neurogenetics 4:309–314. doi: 10.3109/01677068709102351,PMID: 3126282

Burmeister SS, Fernald RD. 2005. Evolutionary conservation of the egr-1 immediate-early gene response in ateleost. The Journal of Comparative Neurology 481:220–232. doi: 10.1002/cne.20380, PMID: 15562507

Cao G, Platisa J, Pieribone VA, Raccuglia D, Kunst M, Nitabach MN. 2013. Genetically targeted opticalelectrophysiology in intact neural circuits. Cell 154:904–913. doi: 10.1016/j.cell.2013.07.027, PMID: 23932121

Dawydow A, Gueta R, Ljaschenko D, Ullrich S, Hermann M, Ehmann N, Gao S, Fiala A, Langenhan T, Nagel G,Kittel RJ. 2014. Channelrhodopsin-2-XXL, a powerful optogenetic tool for low-light applications. PNAS 111:13972–13977. doi: 10.1073/pnas.1408269111, PMID: 25201989

Edelstein A, Amodaj N, Hoover K, Vale R, Stuurman N. 2010. Computer control of microscopes using mManager.Current Protocols in Molecular Biology 14. doi: 10.1002/0471142727.mb1420s92, PMID: 20890901

Eden E, Lipson D, Yogev S, Yakhini Z. 2007. Discovering motifs in ranked lists of DNA sequences. PLoSComputational Biology 3:e39. doi: 10.1371/journal.pcbi.0030039, PMID: 17381235

Eden E, Navon R, Steinfeld I, Lipson D, Yakhini Z. 2009. GOrilla: a tool for discovery and visualization of enrichedGO terms in ranked gene lists. BMC Bioinformatics 10:48. doi: 10.1186/1471-2105-10-48, PMID: 19192299

Fields RD, Eshete F, Stevens B, Itoh K. 1997. Action potential-dependent regulation of gene expression:temporal specificity in ca2+, cAMP-responsive element binding proteins, and mitogen-activated protein kinasesignaling. Journal of Neuroscience 17:7252–7266. PMID: 9295372

Fowler T, Sen R, Roy AL. 2011. Regulation of primary response genes. Molecular Cell 44:348–360. doi: 10.1016/j.molcel.2011.09.014, PMID: 22055182

Fujita N, Nagata Y, Nishiuchi T, Sato M, Iwami M, Kiya T. 2013. Visualization of neural activity in insect brainsusing a conserved immediate early gene, Hr38. Current Biology 23:2063–2070. doi: 10.1016/j.cub.2013.08.051,PMID: 24120640

Greenberg ME, Ziff EB, Greene LA. 1986. Stimulation of neuronal acetylcholine receptors induces rapid genetranscription. Science 234:80–83. doi: 10.1126/science.3749894, PMID: 3749894

Guan Z, Saraswati S, Adolfsen B, Littleton JT. 2005. Genome-wide transcriptional changes associated withenhanced activity in the drosophila nervous system. Neuron 48:91–107. doi: 10.1016/j.neuron.2005.08.036,PMID: 16202711

Guo F, Cerullo I, Chen X, Rosbash M. 2014. PDF neuron firing phase-shifts key circadian activity neurons inDrosophila. eLife 3:e02780. doi: 10.7554/eLife.02780

Guo F, Yu J, Jung HJ, Abruzzi KC, Luo W, Griffith LC, Rosbash M. 2016. Circadian neuron feedback controls thedrosophila sleep–activity profile. Nature 536:292–297. doi: 10.1038/nature19097, PMID: 27479324

Haynes PR, Christmann BL, Griffith LC. 2015. A single pair of neurons links sleep to memory consolidation inDrosophila melanogaster. eLife 4:e03868. doi: 10.7554/eLife.03868

Helfrich-Forster C. 1997. Development of pigment-dispersing hormone-immunoreactive neurons in the nervoussystem of drosophila melanogaster. The Journal of Comparative Neurology 380:335–354. doi: 10.1002/(SICI)1096-9861(19970414)380:3<335::AID-CNE4>3.3.CO;2-M

Herschman HR. 1991. Primary response genes induced by growth factors and tumor promoters. Annual Reviewof Biochemistry 60:281–319. doi: 10.1146/annurev.bi.60.070191.001433, PMID: 1883198

Hong EJ, McCord AE, Greenberg ME. 2008. A biological function for the neuronal activity-dependentcomponent of bdnf transcription in the development of cortical inhibition. Neuron 60:610–624. doi: 10.1016/j.neuron.2008.09.024, PMID: 19038219

Huh GS, Boulanger LM, Du H, Riquelme PA, Brotz TM, Shatz CJ. 2000. Functional requirement for class I MHC inCNS development and plasticity. Science 290:2155–2159. doi: 10.1126/science.290.5499.2155, PMID: 11118151

Jones SD, Hall DJ, Rollins BJ, Stiles CD. 1988. Platelet-derived growth factor generates at least two distinctintracellular signals that modulate gene expression. Cold Spring Harbor Symposia on Quantitative Biology 53:531–536. doi: 10.1101/SQB.1988.053.01.061, PMID: 3254779

Kandel ER. 2001. The molecular biology of memory storage: a dialog between genes and synapses.Bioscience Reports 21:565–611. doi: 10.1023/A:1014775008533, PMID: 12168768

Kim TK, Hemberg M, Gray JM, Costa AM, Bear DM, Wu J, Harmin DA, Laptewicz M, Barbara-Haley K, KuerstenS, Markenscoff-Papadimitriou E, Kuhl D, Bito H, Worley PF, Kreiman G, Greenberg ME. 2010. Widespreadtranscription at neuronal activity-regulated enhancers. Nature 465:182–187. doi: 10.1038/nature09033,PMID: 20393465

Kiya T, Kunieda T, Kubo T. 2007. Increased neural activity of a mushroom body neuron subtype in the brains offorager honeybees. PLoS One 2:e371. doi: 10.1371/journal.pone.0000371, PMID: 17440607

Klapoetke NC, Murata Y, Kim SS, Pulver SR, Birdsey-Benson A, Cho YK, Morimoto TK, Chuong AS, CarpenterEJ, Tian Z, Wang J, Xie Y, Yan Z, Zhang Y, Chow BY, Surek B, Melkonian M, Jayaraman V, Constantine-PatonM, Wong GK, et al. 2014. Independent optical excitation of distinct neural populations. Nature Methods 11:338–346. doi: 10.1038/nmeth.2836, PMID: 24509633

Kula-Eversole E, Nagoshi E, Shang Y, Rodriguez J, Allada R, Rosbash M. 2010. Surprising gene expressionpatterns within and between PDF-containing circadian neurons in drosophila. PNAS 107:13497–13502. doi: 10.1073/pnas.1002081107, PMID: 20624977

Langmead B, Salzberg SL. 2012. Fast gapped-read alignment with bowtie 2. Nature Methods 9:357–359.doi: 10.1038/nmeth.1923, PMID: 22388286

Chen et al. eLife 2016;5:e19942. DOI: 10.7554/eLife.19942 19 of 21

Research article Neuroscience

Liang X, Holy TE, Taghert PH. 2016. Synchronous drosophila circadian pacemakers display nonsynchronous Ca2+

rhythms in vivo. Science 351:976–981. doi: 10.1126/science.aad3997, PMID: 26917772Lin Y, Bloodgood BL, Hauser JL, Lapan AD, Koon AC, Kim TK, Hu LS, Malik AN, Greenberg ME. 2008. Activity-dependent regulation of inhibitory synapse development by Npas4. Nature 455:1198–1204. doi: 10.1038/nature07319, PMID: 18815592

Lopes JS, Abril-de-Abreu R, Oliveira RF. 2015. Brain transcriptomic response to social eavesdropping in zebrafish(Danio rerio). PLoS One 10:e0145801. doi: 10.1371/journal.pone.0145801, PMID: 26713440

Lutz CC, Robinson GE. 2013. Activity-dependent gene expression in honey bee mushroom bodies in response toorientation flight. Journal of Experimental Biology 216:2031–2038. doi: 10.1242/jeb.084905, PMID: 23678099

Mardinly AR, Spiegel I, Patrizi A, Centofante E, Bazinet JE, Tzeng CP, Mandel-Brehm C, Harmin DA, Adesnik H,Fagiolini M, Greenberg ME. 2016. Sensory experience regulates cortical inhibition by inducing IGF1 in VIPneurons. Nature 531:371–375. doi: 10.1038/nature17187, PMID: 26958833

McCarthy DJ, Chen Y, Smyth GK. 2012. Differential expression analysis of multifactor RNA-Seq experiments withrespect to biological variation. Nucleic Acids Research 40:4288–4297. doi: 10.1093/nar/gks042, PMID: 22287627

Menet JS, Pescatore S, Rosbash M. 2014. CLOCK:BMAL1 is a pioneer-like transcription factor. Genes &Development 28:8–13. doi: 10.1101/gad.228536.113, PMID: 24395244

Mokin M, Keifer J. 2005. Expression of the immediate-early gene-encoded protein Egr-1 (zif268) during in vitroclassical conditioning. Learning & Memory 12:144–149. doi: 10.1101/lm.87305, PMID: 15805312

Moore HA, Whitmore D. 2014. Circadian rhythmicity and light sensitivity of the zebrafish brain. PLoS One 9:e86176. doi: 10.1371/journal.pone.0086176, PMID: 24465943

Nagoshi E, Sugino K, Kula E, Okazaki E, Tachibana T, Nelson S, Rosbash M. 2010. Dissecting differential geneexpression within the circadian neuronal circuit of drosophila. Nature Neuroscience 13:60–68. doi: 10.1038/nn.2451, PMID: 19966839

Nestler EJ. 2001. Molecular basis of long-term plasticity underlying addiction. Nature Reviews Neuroscience 2:119–128. doi: 10.1038/35053570, PMID: 11252991

Petesch SJ, Lis JT. 2008. Rapid, transcription-independent loss of nucleosomes over a large chromatin domain atHsp70 loci. Cell 134:74–84. doi: 10.1016/j.cell.2008.05.029, PMID: 18614012

Robinson MD, McCarthy DJ, Smyth GK. 2010. edgeR: a Bioconductor package for differential expressionanalysis of digital gene expression data. Bioinformatics 26:139–140. doi: 10.1093/bioinformatics/btp616,PMID: 19910308

Robinson MD, Smyth GK. 2007. Moderated statistical tests for assessing differences in tag abundance.Bioinformatics 23:2881–2887. doi: 10.1093/bioinformatics/btm453, PMID: 17881408

Robinson MD, Smyth GK. 2008. Small-sample estimation of negative binomial dispersion, with applications toSAGE data. Biostatistics 9:321–332. doi: 10.1093/biostatistics/kxm030, PMID: 17728317

Rodriguez J, Tang CH, Khodor YL, Vodala S, Menet JS, Rosbash M. 2013. Nascent-Seq analysis of drosophilacycling gene expression. PNAS 110:E275–E284. doi: 10.1073/pnas.1219969110, PMID: 23297234

Shang Y, Donelson NC, Vecsey CG, Guo F, Rosbash M, Griffith LC. 2013. Short neuropeptide F is a sleep-promoting inhibitory modulator. Neuron 80:171–183. doi: 10.1016/j.neuron.2013.07.029, PMID: 24094110

Spiegel I, Mardinly AR, Gabel HW, Bazinet JE, Couch CH, Tzeng CP, Harmin DA, Greenberg ME. 2014. Npas4regulates excitatory-inhibitory balance within neural circuits through cell-type-specific gene programs. Cell 157:1216–1229. doi: 10.1016/j.cell.2014.03.058, PMID: 24855953

Spitzer NC, Lautermilch NJ, Smith RD, Gomez TM. 2000. Coding of neuronal differentiation by calciumtransients. BioEssays 22:811–817. doi: 10.1002/1521-1878(200009)22:9<811::AID-BIES6>3.0.CO;2-G, PMID: 10944583

Stoleru D, Peng Y, Agosto J, Rosbash M. 2004. Coupled oscillators control morning and evening locomotorbehaviour of drosophila. Nature 431:862–868. doi: 10.1038/nature02926, PMID: 15483615

Tanenhaus AK, Zhang J, Yin JC. 2012. In vivo circadian oscillation of dCREB2 and NF-kB activity in thedrosophila nervous system. PLoS One 7:e45130. doi: 10.1371/journal.pone.0045130, PMID: 23077489

Trapnell C, Pachter L, Salzberg SL. 2009. TopHat: discovering splice junctions with RNA-Seq. Bioinformatics 25:1105–1111. doi: 10.1093/bioinformatics/btp120, PMID: 19289445

Trapnell C, Roberts A, Goff L, Pertea G, Kim D, Kelley DR, Pimentel H, Salzberg SL, Rinn JL, Pachter L. 2012.Differential gene and transcript expression analysis of RNA-seq experiments with TopHat and cufflinks. NatureProtocols 7:562–578. doi: 10.1038/nprot.2012.016, PMID: 22383036

Tullai JW, Schaffer ME, Mullenbrock S, Sholder G, Kasif S, Cooper GM. 2007. Immediate-early and delayedprimary response genes are distinct in function and genomic architecture. Journal of Biological Chemistry 282:23981–23995. doi: 10.1074/jbc.M702044200, PMID: 17575275

Ueno T, Tomita J, Tanimoto H, Endo K, Ito K, Kume S, Kume K. 2012. Identification of a dopamine pathway thatregulates sleep and arousal in drosophila. Nature Neuroscience 15:1516–1523. doi: 10.1038/nn.3238,PMID: 23064381

Wang JW, Wong AM, Flores J, Vosshall LB, Axel R. 2003. Two-photon calcium imaging reveals an odor-evokedmap of activity in the fly brain. Cell 112:271–282. doi: 10.1016/S0092-8674(03)00004-7, PMID: 12553914

West AE, Greenberg ME. 2011. Neuronal activity-regulated gene transcription in synapse development andcognitive function. Cold Spring Harbor Perspectives in Biology 3:a005744. doi: 10.1101/cshperspect.a005744,PMID: 21555405

Wu C. 1997. Chromatin remodeling and the control of gene expression. Journal of Biological Chemistry 272:28171–28174. doi: 10.1074/jbc.272.45.28171, PMID: 9353261

Chen et al. eLife 2016;5:e19942. DOI: 10.7554/eLife.19942 20 of 21

Research article Neuroscience

Zhou X, Lindsay H, Robinson MD. 2014. Robustly detecting differential expression in RNA sequencing data usingobservation weights. Nucleic Acids Research 42:e91. doi: 10.1093/nar/gku310, PMID: 24753412

Chen et al. eLife 2016;5:e19942. DOI: 10.7554/eLife.19942 21 of 21

Research article Neuroscience